lecture 1
play

LECTURE 1: ubiquity ; INTRODUCTION interconnection ; intelligence - PDF document

Overview Five ongoing trends have marked the history of computing: LECTURE 1: ubiquity ; INTRODUCTION interconnection ; intelligence ; delegation ; and Multiagent Systems human-orientation Based on An Introduction to


  1. Overview � Five ongoing trends have marked the history of computing: LECTURE 1: � ubiquity ; INTRODUCTION � interconnection ; � intelligence ; � delegation ; and Multiagent Systems � human-orientation Based on “An Introduction to MultiAgent Systems” by Michael Wooldridge, John Wiley & Sons, 2002. http://www.csc.liv.ac.uk/˜mjw/pubs/imas/ Ubiquity Interconnection � Computer systems today no longer stand � The continual reduction in cost of computing alone, but are networked into large capability has made it possible to introduce distributed systems processing power into places and devices � The internet is an obvious example, but that would have once been uneconomic networking is spreading its ever-growing � As processing capability spreads, tentacles… sophistication (and intelligence of a sort) � Since distributed and concurrent systems becomes ubiquitous have become the norm, some researchers � What could benefit from having a processor are putting forward theoretical models that embedded in it…? portray computing as primarily a process of interaction Intelligence Delegation � Computers are doing more for us – without � The complexity of tasks that we are capable our intervention of automating and delegating to computers � We are giving control to computers, even in has grown steadily safety critical tasks � If you don’t feel comfortable with this � One example: fly-by-wire aircraft, where the definition of “intelligence”, it’s probably machine’s judgment may be trusted more because you are a human than an experienced pilot � Next on the agenda: fly-by-wire cars, intelligent braking systems, cruise control that maintains distance from car in front… 1

  2. Human Orientation Programming progression… � Programming has progressed through: � The movement away from machine-oriented � machine code; views of programming toward concepts and � assembly language; metaphors that more closely reflect the way we ourselves understand the world � machine-independent programming languages; � sub-routines; � Programmers (and users!) relate to the � procedures & functions; machine differently � abstract data types; � Programmers conceptualize and implement � objects; software in terms of higher-level – more human-oriented – abstractions to agents . Global Computing Where does it bring us? � What techniques might be needed to deal � Delegation and Intelligence imply the need to with systems composed of 10 10 processors? build computer systems that can act effectively on our behalf � Don’t be deterred by its seeming to be “science fiction” � This implies: � The ability of computer systems to act � Hundreds of millions of people connected by independently email once seemed to be “science fiction”… � The ability of computer systems to act in a way � Let’s assume that current software that represents our best interests while interacting development models can’t handle this… with other humans or systems Interconnection and Distribution So Computer Science expands… � Interconnection and Distribution have � These issues were not studied in Computer become core motifs in Computer Science Science until recently � But Interconnection and Distribution, coupled � All of these trends have led to the emergence with the need for systems to represent our of a new field in Computer Science: best interests, implies systems that can multiagent systems cooperate and reach agreements (or even compete ) with other systems that have different interests (much as we do with other people) 2

  3. Agents, a Definition Multiagent Systems, a Definition � A multiagent system is one that consists � An agent is a computer system that is of a number of agents, which interact with capable of independent action on behalf of one-another its user or owner (figuring out what needs � In the most general case, agents will be to be done to satisfy design objectives, acting on behalf of users with different rather than constantly being told) goals and motivations � To successfully interact, they will require the ability to cooperate , coordinate , and negotiate with each other, much as people do Agent Design, Society Design Multiagent Systems � The course covers two key problems: � In Multiagent Systems, we address questions such as: � How do we build agents capable of independent, autonomous action, so that they can successfully carry � How can cooperation emerge in societies of self- out tasks we delegate to them? interested agents? � How do we build agents that are capable of interacting � What kinds of languages can agents use to (cooperating, coordinating, negotiating) with other communicate? agents in order to successfully carry out those � How can self-interested agents recognize conflict, delegated tasks, especially when the other agents and how can they (nevertheless) reach cannot be assumed to share the same interests/goals? agreement? � The first problem is agent design , the second is � How can autonomous agents coordinate their society design (micro/macro) activities so as to cooperatively achieve goals? Multiagent Systems The Vision Thing � It’s easiest to understand the field of multiagent � While these questions are all addressed systems if you understand researchers’ vision of in part by other disciplines (notably the future economics and social sciences), what � Fortunately, different researchers have different makes the multiagent systems field visions unique is that it emphasizes that the � The amalgamation of these visions (and agents in question are computational, research directions, and methodologies, and information processing entities. interests, and…) define the field � But the field’s researchers clearly have enough in common to consider each other’s work relevant to their own 3

  4. Spacecraft Control Deep Space 1 � http://nmp.jpl.nasa.gov/ds1/ � When a space probe makes its long flight from Earth � “Deep Space 1 to the outer planets, a ground crew is usually launched from Cape required to continually track its progress, and decide Canaveral on October 24, how to deal with unexpected eventualities. This is 1998. During a highly costly and, if decisions are required quickly , it is successful primary mission, it tested 12 advanced, high-risk technologies in simply not practicable. For these reasons, space. In an extremely successful extended organizations like NASA are seriously investigating mission, it encountered comet Borrelly and the possibility of making probes more autonomous returned the best images and other science data — giving them richer decision making capabilities ever from a comet. During its fully successful hyperextended mission, it conducted further and responsibilities. technology tests. The spacecraft was retired on � This is not fiction: NASA ’s DS1 has done it! December 18, 2001.” – NASA Web site Autonomous Agents for specialized tasks Air Traffic Control � “A key air-traffic control system…suddenly � The DS1 example is one of a generic class fails, leaving flights in the vicinity of the airport � Agents (and their physical instantiation in with no air-traffic control support. Fortunately, robots) have a role to play in high-risk autonomous air-traffic control systems in situations, unsuitable or impossible for nearby airports recognize the failure of their humans peer, and cooperate to track and deal with all � The degree of autonomy will differ depending affected flights.” on the situation (remote human control may � Systems taking the initiative when necessary be an alternative, but not always) � Agents cooperating to solve problems beyond the capabilities of any individual agent Internet Agents What if the agents become better? � Searching the Internet for the answer to a � Internet agents need not simply search specific query can be a long and tedious � They can plan, arrange, buy, negotiate – process. So, why not allow a computer program carry out arrangements of all sorts that would — an agent — do searches for us? The agent normally be done by their human user would typically be given a query that would � As more can be done electronically, software require synthesizing pieces of information from agents theoretically have more access to various different Internet information sources. systems that affect the real-world Failure would occur when a particular resource � But new research problems arise just as was unavailable, (perhaps due to network quickly… failure), or where results could not be obtained. 4

Recommend


More recommend